5014559
doi
10.5281/zenodo.5014559
oai:zenodo.org:5014559
user-ai4media
Zhun Zhong
University of Trento, Italy
Zhiming Luo
Xiamen University, China
Yuanzheng Cai
Minjiang University, China
Yaojin Lin
Minnan Normal University, China
Shaozi Li
Xiamen University, China
Nicu Sebe
University of Trento, Italy
Joint Noise-Tolerant Learning and Meta Camera Shift Adaptation for Unsupervised Person Re-Identification
Fengxiang Yang
Xiamen University, China
info:eu-repo/semantics/openAccess
Creative Commons Attribution 4.0 International
https://creativecommons.org/licenses/by/4.0/legalcode
<p>This paper considers the problem of unsupervised person re-identification (re-ID), which aims to learn discriminative models with unlabeled data. One popular method is to obtain pseudo-label by clustering and use them to optimize the model. Although this kind of approach has shown promising accuracy, it is hampered by 1) noisy labels produced by clustering and 2) feature variations caused by camera shift. The former will lead to incorrect optimization and thus hinders the model accuracy. The latter will result in assigning the intra-class samples of different cameras to different pseudo-label, making the model sensitive to camera variations. In this paper, we propose a unified framework to solve both problems. Concretely, we propose a Dynamic and Symmetric Cross Entropy loss (DSCE) to deal with noisy samples and a camera-aware meta-learning algorithm (MetaCam) to adapt camera shift. DSCE can alleviate the negative effects of noisy samples and accommodate the change of clusters after each clustering step. MetaCam simulates cross-camera constraint by splitting the training data into meta-train and meta-test based on camera IDs.<br>
With the interacted gradient from meta-train and meta-test, the model is enforced to learn camera-invariant features. Extensive experiments on three re-ID benchmarks show the effectiveness and the complementary of the proposed DSCE and MetaCam. Our method outperforms the state-of-the-art methods on both fully unsupervised re-ID and unsupervised domain adaptive re-ID.</p>
Zenodo
2021-06-22
info:eu-repo/semantics/conferencePaper
5014558
user-ai4media
award_title=A European Excellence Centre for Media, Society and Democracy; award_number=951911; award_identifiers_scheme=url; award_identifiers_identifier=https://cordis.europa.eu/projects/951911; funder_id=00k4n6c32; funder_name=European Commission;
1624412899.437216
851512
md5:65414290237b9d4b68e8f73827d21651
https://zenodo.org/records/5014559/files/Yang_Joint_Noise-Tolerant_Learning_and_Meta_Camera_Shift_Adaptation_for_Unsupervised_CVPR_2021_paper.pdf
public
10.5281/zenodo.5014558
isVersionOf
doi